I'm Anti "AI"

As we all learned from The Good Place, the modern world is complicated, and our actions have unintended consequences.

A big unintended consequence of using automated image and text generation technology is that the data centers that process these requests waste huge amounts of energy and water.

Where we can, I think it is good for all of us to minimize our carbon footprint. (I even add "-ai" after my Google searches so that they don't automatically waste additional resources!)

Tools for automatic media generation require large, well-annotated sets of training data. The companies that make these tools have stolen their training data from authors and artists. They collected it without their consent and without compensating them. They then try to sell these tools as a way to replace authors and artists under the guise of efficiency. They are trying to automate what people enjoy and make a living doing.

In addition, training these tools requires humans who can test and determine their limits, which involves getting them to create violent and disturbing content. These are low-paying jobs. People are not adequately recognized or supported for their work training these tools. 

A tool like ChatGPT is designed to make things up. It doesn't search for information, and it doesn't think. It has no understanding of the world. The only thing it does is use mathematical formulas to generate words based on probability. It might get a lot of things "right", but unless you are able to verify for yourself, you can't trust that it will get everything (or anything) right.

There are lots of examples of this. I have posted one here that is very ironic: a suggested summer reading list was made with automatic text generation and it named several books that don't exist. There are also examples of academic papers with fake citations, and of automatically generated recommendations that would cause injury if they were followed. In all these cases, the tool is doing exactly what it is designed to do: making things up based only on probable word combinations. The results can't be trusted.

If you can put aside the environmental and ethical issues, it might be fun to think of an automatic text generation tool like a Magic 8 ball, a tarot spread, or your horoscope. They are words/images for us to think about, but they aren't based in truth or fact.

When I was a student taking Statistics, we had to learn how to do things like calculate a standard deviation by hand, even though we all had calculators that could do it for us. The reason was so that we could understand what the logic behind a standard deviation is, and so that we would know what the tool was doing for us.

I'm a professor of linguistics. I like reading and writing, and I am in awe of what humans can do with language. As an academic, I value knowing where information comes from. As somebody who lives in the US, I see that people are very easily manipulated when they cannot evaluate the source for information they get.

All of these factors make me opposed to the use of automatic text generation in educational and academic contexts. Reading and writing are valuable skills that require practice. We cannot rely on automatic text generators when reading and writing are the skills we seek to develop.